Binaural bark subband pre-processing of nonstationary signals for noise robust speech feature extraction.
Mike PetersPublished in: ICASSP (1999)
Keyphrases
- non stationary
- subband
- feature extraction
- noisy environments
- preprocessing
- wavelet transform
- speech signal
- feature vectors
- frequency domain
- bandpass
- low frequency
- filter bank
- high frequency
- multiresolution
- image compression
- wavelet coefficients
- wavelet decomposition
- wavelet packet
- empirical mode decomposition
- random fields
- discrete wavelet transform
- speech recognition
- low pass
- blind source separation
- wavelet domain
- bit rate
- subband coding
- noise reduction
- sound source
- high pass
- image coding
- frequency band
- signal processing
- additive noise
- image coder
- instantaneous frequency
- concept drift
- computationally efficient
- autoregressive
- gabor filters
- feature space
- multiscale
- face recognition
- transform coefficients
- similarity measure
- feature selection